Flatten nested JSON (has multiple list) into multiple pandas dataframe columns-CodePudding

I have a pandas column with nested json data string. I'd like to flatten the data into multiple pandas columns. I have data like this:

{
 'A': '123',
 'B': '2019-08-26', 
 'C': [
       {
        'a': 'stop', 
        'b': 'A '
       },
       {
        'a': 'go', 
        'b': 'C '
       }
      ], 
'D': [],
'E': [
      {
       'a': 'Don', 
      'b': 1
      },
      {
       'b': 12
      }
     ], 
}

For each cell in pandas column, I'd like parse this string and create multiple columns. Expected output looks something like this:

| A | B | C.a | C.b | D.a | D.b | E.a | E.b |
|---- |------|-----|-----|-----|-----|-----|-----|
| 123  | 2019-08-26  | stop | A  | Nan | Nan | Don | 1 |
| 123  | 2019-08-26  | go | C  | Nan | Nan | Don | 1 |
| 123  | 2019-08-26  | stop | A  | Nan | Nan | NaN | 12 |
| 123  | 2019-08-26  | go | C  | Nan | Nan | Nan | 12 |

I tried using json_normalize, but it return error.... Please help me :(

CodePudding user response：

Use pd.json_normalize with df.explode and pd.concat:

In [308]: x = pd.json_normalize(j).explode('C').explode('E')  
In [310]: r = pd.concat([x.drop(['C', 'E'], 1).reset_index(drop=True), pd.json_normalize(x.C), pd.json_normalize(x.E)], 1)

In [316]: C_cols = [f'C.{i}' for i in pd.json_normalize(x.C).columns]    
In [317]: E_cols = [f'E.{i}' for i in pd.json_normalize(x.E).columns]

In [323]: r.columns = [*x.drop(['C', 'E'], 1).columns , *C_cols, *E_cols]

In [324]: r
Out[324]: 
     A           B   D   C.a C.b  E.a  E.b
0  123  2019-08-26  []  stop  A   Don    1
1  123  2019-08-26  []  stop  A   NaN   12
2  123  2019-08-26  []    go  C   Don    1
3  123  2019-08-26  []    go  C   NaN   12

CodePudding user response：

Similar to @Mayank Porwal's answer, first use pd.json_normalize df.explode. Then use str.get method to collect the values from dictionaries in columns ['C','D','E']:

df = pd.json_normalize(json_data).explode('C').explode('E')
for col in ['C','D','E']:
    for i in ['a','b']:
        df[col '.' i] = df[col].str.get(i)
df['E.a'].replace({None:np.nan}, inplace=True)
df = df.drop(['C','E','D'], axis=1).sort_values(by='E.b')

Output:

     A           B   C.a C.b  D.a  D.b  E.a  E.b
0  123  2019-08-26  stop  A   NaN  NaN  Don    1
0  123  2019-08-26    go  C   NaN  NaN  Don    1
0  123  2019-08-26  stop  A   NaN  NaN  NaN   12
0  123  2019-08-26    go  C   NaN  NaN  NaN   12